Benchmarking Strategy for Arabic Screen-Rendered Word Recognition

Abstract

This chapter presents a new benchmarking strategy for Arabic screen-based word recognition. Firstly, we report on the creation of the new APTI (Arabic Printed Text Image) database. This database is a large-scale benchmarking of open-vocabulary, multi-font, multi-size and multi-style word recognition systems in Arabic. Such systems take as input a text image and compute as output a character string corresponding to the text included in the image. The challenges that are addressed by the database are in the variability of the sizes, fonts and styles used to generate the images. A focus is also given on low resolution images where anti-aliasing is generating noise on the characters being recognized. The database contains 45,313,600 single word images totalling more than 250 million characters. Ground truth annotation is provided for each image from an XML file. The annotation includes the number of characters, the number of pieces of Arabic words (PAWs), the sequence of characters, the size, the style, the font used to generate each image, etc. Secondly, we describe the Arabic Recognition Competition: Multi-Font Multi-Size Digitally Represented Text held in the context of the 11th International Conference on Document Analysis and Recognition (ICDAR’2011), during September 18–21, 2011, Beijing, China. This first edition of the competition used the freely available APTI database. Two groups with three systems participated in the competition. The systems were compared using the recognition rates at the character and word levels. The systems were tested on one test dataset which is unknown to all participants (set 6 of APTI database). The systems were compared on the ground of the most important characteristic of classification systems: the recognition rate. A short description of the participating groups, their systems, the experimental setup and the observed results are presented. Thirdly, we present our DIVA-REGIM system (out of competition at ICDAR’2011) with all results of the Arabic recognition competition protocols.